Claude Code is Anthropic’s coding assistant for software projects. It is included with eligible paid Claude plans. You can also use it with metered API billing, which charges for the text the model processes and generates. Choose based on how often you use Claude Code: subscriptions have a set price and usage limits, while API billing charges for the tokens you use.
Pro and Max plans have monthly fees, while API costs depend on the model and workload. Claude Code shares subscription usage with Claude chat, and five-hour and weekly limits apply. The monthly fee does not include unlimited coding time.
Claude Code plan prices
Paid individual plans include Claude Code. Current plan summaries also list it as included with eligible Team plans. Claude Code has no separate monthly fee.
| Plan | Listed price | Claude Code usage |
|---|---|---|
| Pro | Monthly billing or $200 yearly | Included, with the standard paid-plan allowance |
| Max 5x | $100 monthly | About five times Pro’s per-session capacity |
| Max (higher-usage tier) | $200 monthly | Higher per-session capacity than Pro |
| Team Standard | Varies by billing terms | Listed as included; confirm the seat’s entitlement before purchase |
| Team Premium | $100–$125 per seat monthly | Listed as included, with about five times Standard usage |
The Claude Code plan breakdown lists these prices and compares the plans. Team prices depend on billing terms. Published summaries disagree about whether Standard seats include Claude Code, so check Anthropic’s current plan details before buying seats.
Max is for people who use Claude Code regularly or for long sessions. Its usage labels compare each plan’s capacity per session; they do not set a fixed message count. Actual use depends on how much text each task sends and receives.
How subscription limits work
Paid-plan usage is shared between Claude Code and Claude chat. Time spent chatting in Claude reduces the allowance available for coding. Usage resets in rolling five-hour windows, with weekly limits also applying. The exact amount available depends on the plan and usage; long conversations and large code changes consume more than short requests.
Message counts alone do not show how long a plan will last. Claude Code sends project context along with your instructions, and long coding tasks use more text. Usage-limit analysis also describes peak-hour throttling, which affects how quickly a session proceeds.
When a subscription limit stops a task, wait for the allowance to reset or continue with separately billed usage credits. Disable usage credits to avoid pay-as-you-go charges. If you want to use only your subscription, check that ANTHROPIC_API_KEY is not set in your shell. Claude Code can use that key to bill API usage instead. The Claude Code pricing overview describes these billing options.
API costs depend on model and token use
With API billing, the provider charges separately for input tokens and output tokens. Input tokens are the text sent to the model; output tokens are the text it generates. A token is a small unit of text. Code, project context, and instructions all count toward the total.
A third-party rate card dated October 2026 lists these standard prices per million tokens:
| Model | Input | Output |
|---|---|---|
| Claude Haiku 4.5 | $1 | $5 |
| Claude Sonnet 5 | $2 | $10 |
| Claude Sonnet 5.5 | $2 | $10 |
| Claude Opus 5 | $5 | $25 |
| Claude Opus 5.5 | $4 | See current pricing |
These list prices are dated. Check the current model pricing table before estimating a bill.
Prompt caching can reduce the price of repeated context. The rate card charges less to read cached input than to send the full text again, and lists separate rates for cache writes. Your bill depends on the model, the amount of input and output, and how much repeated context qualifies for caching.
Coding tasks use different amounts of tokens. A small question needs much less context than a repository-wide refactor or a long-running agent task. Prompt counts alone do not capture input and output token totals. Estimate costs by tracking those tokens for representative work, then applying the model’s current rates.
API limits and billing controls
API billing has limits separate from subscription allowances. Anthropic sets a per-minute request limit, along with separate per-minute limits for input and output tokens. These limits depend on the model and account tier. The official API rate-limit documentation explains the measures and available tiers.
The documented monthly spend caps are $500 for Start, $1,000 for Build, and $200,000 for Scale. When you reach a cap, API use pauses until the next month. Anthropic’s rate-limit policy says higher limits depend on account tier and usage.
Before running repeated coding tasks through the API, set an organization-level spend cap and review usage. A rate limit controls how quickly requests run. A spend cap sets the maximum monthly bill. Both help if a script or agent repeats requests unexpectedly.
Choosing between a subscription and API billing
API billing suits occasional use because you pay for the tokens you use. A subscription offers more predictable costs for frequent interactive coding, especially when API list prices for the same workload exceed the monthly plan fee. Heavy users should compare their token use and model choice instead of relying on message counts.
Choose Pro for regular, moderate Claude use. Consider Max if frequent or long Claude Code sessions keep hitting Pro’s limits. Choose API billing for occasional tasks or controlled automation, and set a spend cap before running requests unattended. If tasks stop mid-session, consider both the five-hour window and weekly allowance before taking on more work.